Proteome-Scale Screen Expands Known Transcriptional Activators

A team out of University of Toronto has greatly expanded the number of known transcriptional activators—proteins that induce transcription of genes into RNA messages—from a handful to around 250, and established how these proteins combine with other cellular machineries to turn genes on. The proteome-scale study also led to new insights into how protein misregulation can lead to cancer.

“This study was a classic fishing expedition where we did not know what we were going to find,” said Mikko Taipale, an associate professor of molecular genetics in the Donnelly Centre for Cellular and Biomolecular Research at the Temerty Faculty of Medicine. “Grant reviewers typically frown upon research that is not hypothesis driven, but that’s the beauty of proteomics. It allows you to cast a net in an unbiased way, and we have found some interesting stuff. We now have a better understanding of which proteins are very strong activators. And we can begin to understand the mechanisms by which they activate transcription.”

To find the activators, the researchers tested the majority of 20,000 human proteins for their ability to activate gene expression in human cells. Many activators were transcription factors (TFs), which directly bind DNA and turn on their target genes, whereas others were helper proteins, or co-factors, that bind TFs and activate their targets together. 

They also found that TFs that are highly similar can talk to different co-factors, explaining why two TFs with essentially identical DNA binding specificities can trigger distinct gene expression. “These activators are not activators in all contexts. It could be that in a gene X they activate, but in gene Y they might actually repress,” Taipale said.

Transcriptional activation occurs through the interaction of the so-called transactivation domains, which are present in the TFs, with the activators. Since the sequences of activation domains are not conserved, they can't be pinpointed by computational methods. For that reason, the team resorted to chopping up 75 activators into pieces and tested the ability of each piece to activate transcription. They identified around 40 activation domains this way.

Using AlphaFold, a bioinformatic tool developed for the prediction of protein structures, the team also found interaction interfaces between TFs and their activators. Although AlphaFold was not designed to predict protein-protein interactions, Taipale expects the software to become the standard tool for these kinds of studies to find functional connections between proteins. “This has been previously nearly impossible to do computationally,” he said.

While many of the identified proteins are novel, some of them were previously detected in tumors in which a TF and its helper protein are permanently joined in an oncogenic fusion protein that ends up activating the wrong genes. Piecing together how TFs interact with different activators could accelerate tailored therapies.

One challenge in therapeutics development has been that TFs are not amenable to targeting by small-molecule drugs. “Transcription factors are really hard to target because they often don’t have druggable pockets, but many of the co-activators are enzymes which means they have pockets that can be targeted,” said Taipale. “For example, when you have a cancer fusion of the transcription factor to the co-activator and you understand the co-activator that the transcription factor interacts with, you may be able to target the co-activator to halt cell proliferation.”

The work was published recently in the journal Molecular Cell.